Can a physical server loose memory dynamically?

My system monitoring (sysutils/munin-node ) shows that my physical thinkserver with 64GB of memory is occasionally loosing (and recovering) memory:

losing-memory.png


To rule out any issues related to munin I verified this by also deploying prometheus sysutils/node_exporter which shows the same events:

memory2.png


In the pictures shown above the server recovered memory without a reboot, but most of the time a reboot is required to recover.
In some cases processes get killed due to lack of available memory.

Is there any plausible explanation why this could happen (other then broken memory hardware)?

This server runs FreeBSD 12.2 but it already happened on FreeBSD 12.1.
This server does not run any bhyve VMs.
 
Back
Top